基于三维激光扫描的织物折皱测试及评价方法研究
发布时间：2018-10-15 06:44
【摘要】：随着生活水平的提高,人们对于服装内在及外在质量提出了更高的要求。织物制成服装后,在穿着过程中难免会起皱,影响美观。现有的织物抗皱性测试方法与实际着装时织物折皱情况差异甚大,不能真实评价织物在实际穿着过程中的抗皱能力。因此对织物抗皱性进行客观、准确地测试和评价就显得很重要。针对这一问题,本文提出一种模拟实际穿着起皱的织物抗皱性测试新方法,构建了模拟人体膝部和肘部起皱的装置。并利用三维激光扫描技术和图像处理技术,对模拟装置产生的折皱提取了三维及二维特征参数,对比分析了两种方法的优劣。选取与主观评价相关性较好的特征参数,利用神经网络技术对折皱等级进行了预测,研究内容及研究结果如下:(1)搭建了能够模拟人体膝部及肘部起皱的装置,该装置能够模拟静态及动态起皱,机械化控制起皱时间、次数、角度,模拟起皱的折皱形态与实际穿着起皱折皱形态非常接近。(2)实验设置了 4个影响织物折皱程度的变量,起皱方式、时间、次数,松量。结果表明:在实验参数下,对于抗皱性差织物来说,松量对织物起皱的形态影响比较大,起皱方式、时间、次数对织物起皱的形态影响较小。对于抗皱性较好的织物来说,4个因素对织物起皱的形态几乎没有影响。(3)基于三维激光扫描提取了反映织物折皱程度的特征参数。结果表明:粗糙度、平均偏移量与折皱等级显著负相关,spearman相关系数在都在0.8以上,单位法向量Z方向绝对值的粗糙度、平均偏移量与折皱等级的相关系数要高于高度方向粗糙度、平均偏移量与折皱等级的相关系数。单位法向量Z方向绝对值的的均值与折皱等级的spearman相关系数为0.679。(4)提取了灰度共生矩阵特征参数能量、熵、对比度、相关性的均值和标准差。结果表明:图像像素为100 x 150时,熵均值及标准差与织物折皱等级的spearman相关系数在0.8以上,熵均值及标准差随着折皱等级的增大而减小。(5)将以上三维指标与二维特征相结合,可以提高折皱等级预测回归模型预测精度。三维参数单位法向量Z方向绝对值平均偏移量和二维参数熵均值与折皱等级的多元线性回归方程的精度为87.5%,三维参数单位法向量Z方向绝对值平均偏移量和折皱等级的线性回归方程的精度为81%,精度提高了6.5%。(6)利用RBF神经网络对折皱等级进行训练及预测,折皱等级预测正确率为83.3%,该模型对素色织物折皱等级预测好于印花织物。部分印花织物折皱等级预测与主观评价偏差较大,一些印花图案对折皱有掩盖效果,可能会导致主观评价不一致,在评价部分印花织物时,需要结合主观评价综合评定其折皱等级。
[Abstract]:With the improvement of living standards, people put forward higher requirements for the internal and external quality of clothing. Fabric made of clothing, wearing process will inevitably wrinkle, affect beauty. There is a great difference between the existing testing methods of crease resistance of fabrics and that of the actual clothes, so it is not possible to evaluate the wrinkle resistance of fabrics in the actual wearing process. Therefore, it is very important to test and evaluate fabric wrinkle resistance objectively and accurately. In order to solve this problem, this paper presents a new method to test the wrinkle resistance of fabric, which simulates the actual wrinkling of fabrics, and constructs a device to simulate wrinkling in the knees and elbows of the human body. Using 3D laser scanning technology and image processing technology, the feature parameters of 3D and 2D are extracted from the wrinkle produced by analog device, and the advantages and disadvantages of the two methods are compared and analyzed. The wrinkle grade is predicted by using neural network technology. The research contents and results are as follows: (1) A device which can simulate wrinkling of human knees and elbows is built. This device can simulate static and dynamic wrinkling, mechanization controls wrinkling time, times, angle, and the wrinkle shape of simulation wrinkle is very close to that of actual dress wrinkle. (2) four variables affecting fabric wrinkle degree are set up in the experiment. Wrinkle mode, time, number of times, loose amount. The results show that under the experimental parameters, for the fabric with poor crease resistance, the amount of looseness has a great influence on the wrinkling morphology of the fabric, and the wrinkling mode, time and times have little effect on the wrinkling morphology of the fabric. For the fabric with good crease resistance, four factors have little effect on the wrinkle shape. (3) based on 3D laser scanning, the characteristic parameters reflecting the crease degree of the fabric are extracted. The results show that the roughness, average deviation and crease grade are negatively correlated, the spearman correlation coefficient is above 0. 8, and the roughness of the absolute value in Z direction of unit normal vector. The correlation coefficient between average migration and wrinkle grade is higher than that of height direction roughness, and the correlation coefficient between average migration and crease grade is higher than that of height direction roughness. The average value of the absolute value of the unit normal vector Z and the spearman correlation coefficient of the wrinkle grade are 0.679. (4) the energy, entropy, contrast, correlation mean and standard deviation of the characteristic parameters of the gray level co-occurrence matrix are extracted. The results show that when the image pixel is 100x150, the spearman correlation coefficient between entropy mean and standard deviation and fabric wrinkle grade is more than 0. 8, and the entropy mean and standard deviation decrease with the increase of wrinkle grade. (5) the above three dimensional indexes are combined with two dimensional features. The prediction accuracy of regression model can be improved. The accuracy of the multivariate linear regression equation of the mean deviation of absolute value in Z direction and the mean value of entropy of two-dimensional parameter and wrinkle grade of three-dimensional parameter unit normal vector is 87.5, and the mean deviation and wrinkle of Z-direction absolute value of three-dimensional parameter unit normal vector are 87.5. The accuracy of the linear regression equation of grade is 81 and the accuracy is improved by 6.5. (6) the crease grade is trained and predicted by RBF neural network. The prediction accuracy of crease grade is 83.3%, and the model is better than that of printed fabric in predicting the wrinkle grade of plain color fabric. The prediction of wrinkle grade of some printed fabrics deviates greatly from subjective evaluation, and some printing patterns have the effect of covering up creases, which may lead to inconsistent subjective evaluation. It is necessary to evaluate its wrinkle grade in combination with subjective evaluation.
【学位授予单位】：浙江理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TS101.923
[Abstract]:With the improvement of living standards, people put forward higher requirements for the internal and external quality of clothing. Fabric made of clothing, wearing process will inevitably wrinkle, affect beauty. There is a great difference between the existing testing methods of crease resistance of fabrics and that of the actual clothes, so it is not possible to evaluate the wrinkle resistance of fabrics in the actual wearing process. Therefore, it is very important to test and evaluate fabric wrinkle resistance objectively and accurately. In order to solve this problem, this paper presents a new method to test the wrinkle resistance of fabric, which simulates the actual wrinkling of fabrics, and constructs a device to simulate wrinkling in the knees and elbows of the human body. Using 3D laser scanning technology and image processing technology, the feature parameters of 3D and 2D are extracted from the wrinkle produced by analog device, and the advantages and disadvantages of the two methods are compared and analyzed. The wrinkle grade is predicted by using neural network technology. The research contents and results are as follows: (1) A device which can simulate wrinkling of human knees and elbows is built. This device can simulate static and dynamic wrinkling, mechanization controls wrinkling time, times, angle, and the wrinkle shape of simulation wrinkle is very close to that of actual dress wrinkle. (2) four variables affecting fabric wrinkle degree are set up in the experiment. Wrinkle mode, time, number of times, loose amount. The results show that under the experimental parameters, for the fabric with poor crease resistance, the amount of looseness has a great influence on the wrinkling morphology of the fabric, and the wrinkling mode, time and times have little effect on the wrinkling morphology of the fabric. For the fabric with good crease resistance, four factors have little effect on the wrinkle shape. (3) based on 3D laser scanning, the characteristic parameters reflecting the crease degree of the fabric are extracted. The results show that the roughness, average deviation and crease grade are negatively correlated, the spearman correlation coefficient is above 0. 8, and the roughness of the absolute value in Z direction of unit normal vector. The correlation coefficient between average migration and wrinkle grade is higher than that of height direction roughness, and the correlation coefficient between average migration and crease grade is higher than that of height direction roughness. The average value of the absolute value of the unit normal vector Z and the spearman correlation coefficient of the wrinkle grade are 0.679. (4) the energy, entropy, contrast, correlation mean and standard deviation of the characteristic parameters of the gray level co-occurrence matrix are extracted. The results show that when the image pixel is 100x150, the spearman correlation coefficient between entropy mean and standard deviation and fabric wrinkle grade is more than 0. 8, and the entropy mean and standard deviation decrease with the increase of wrinkle grade. (5) the above three dimensional indexes are combined with two dimensional features. The prediction accuracy of regression model can be improved. The accuracy of the multivariate linear regression equation of the mean deviation of absolute value in Z direction and the mean value of entropy of two-dimensional parameter and wrinkle grade of three-dimensional parameter unit normal vector is 87.5, and the mean deviation and wrinkle of Z-direction absolute value of three-dimensional parameter unit normal vector are 87.5. The accuracy of the linear regression equation of grade is 81 and the accuracy is improved by 6.5. (6) the crease grade is trained and predicted by RBF neural network. The prediction accuracy of crease grade is 83.3%, and the model is better than that of printed fabric in predicting the wrinkle grade of plain color fabric. The prediction of wrinkle grade of some printed fabrics deviates greatly from subjective evaluation, and some printing patterns have the effect of covering up creases, which may lead to inconsistent subjective evaluation. It is necessary to evaluate its wrinkle grade in combination with subjective evaluation.
【学位授予单位】：浙江理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TS101.923
【参考文献】
相关期刊论文 前10条
1 张晓婷;洪剑寒;g癜，
本文编号：2271699
本文链接：https://www.wllwen.com/shoufeilunwen/boshibiyelunwen/2271699.html
